Multi-Layered Data Classification for Biomarker Discovery

• Activity Number: 62
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Samuel Mueller*+ and Ellis Patrick and Jean Yee Hwa Yang
• Companies: University of Sydney and University of Sydney and University of Sydney
• Keywords: data integration ; bioinformatics ; gene expression ; omics data

Abstract

Over the last decade, several statistical techniques have been proposed to tackle genome-wide expression data. However, with the advancement of many other high-throughput biotechnologies, the interest of researchers has been focusing on utilizing multiple data sources together with the clinical data, to improve the prognosis of disease outcome.

Integrating the components from different platforms has become a crucial step to better understand the relationships between clinical and -omics data and the information they provide to classify some response. The statistical task to preserve the stability and interpretability of the classifier has become more challenging in this framework. One major issue is that the large dimension of -omics data can completely dominate the modelling procedure and it is an open question how to best combine different types of variables.

This talk will present our most recent results on improving upon standard classification procedures for Ovarian and Melanoma Cancer Data. We will use a multi-stage process that involves simultaneously changing the number of observations and features to classify well different disease outcome.